The Fraunhofer Institute for Computer Architecture and Software Technology (Fraunhofer FIRST) is involved in a large number of research and industry projects on connecting SOA and grid computing. As a co-founder of the Fraunhofer Grid Alliance (FhGA), Fraunhofer FIRST is one of the key players involved in the establishment of the Fraunhofer Resource Grid (FhRG). This cross-institution grid platform of the Fraunhofer Gesellschaft provides extensive IT resources – hardware, data, software, and services – in an SOA.
The projects on connecting SOA with grid computing focus on the issue of how geographically distributed resources can be effectively combined within a workflow. Among other things, Fraunhofer FIRST developed the Petri net-based workflow service “Grid Workflow Execution Service (GWES)”, which can be used to integrate and execute existing web services, grid services, and conventional program calls easily in a automated workflow (www.gridworkflow.org/gwes). One other important task is to make IT resources available in a service-oriented grid in a user-friendly way. Here, Fraunhofer FIRST is using a number of different concepts. With “Instant-Grid,” the institute produced a Knoppix-based live CD, which turns an existing computer infrastructure into a functioning grid environment in just a few minutes. Partners in this project include the Gesellschaft für wissenschaftliche Datenverarbeitung mbH in Göttingen and ed-media GmbH in Kaiserslautern. Together with users and developers at the German Aerospace Center, a “VirtualLab Platform” was also developed, which enables web- and grid-based working with special software across the whole breadth of scientific and engineering computing. For the connection of SOA and grid computing, security concepts are also required that offer a secure access control for web services and at the same time ensure the confidentiality of user data. In this area, Fraunhofer FIRST is working with the University of Potsdam on the prototype of a fine-granular grid and web service security solution that enables certificate-based access control and a restricted delegation of rights for web services.
ERAMAS: Intelligent disaster management
The advantages of combined grid computing and service-oriented architecture can be explained using three practical examples.
The analysis and management system ERAMAS (Environmental Risk Analysis and Management System) makes it possible to forecast and evaluate the dispersion of carcinogenic and chemically toxic pollutants in the air, soil, and groundwater using high-quality simulation programs to assess the danger for humans. Areas of application for this system include the transportation of hazardous materials, incidents in industrial plants, or terrorist attacks. The analysis is conducted under real-time conditions, which means the computing results must be available quickly enough to enable the emergency services to make decisions on evacuation, for example, and implement these before the pollutants have reached the affected areas.
ERAMAS, developed by Fraunhofer FIRST together with Dresdner Grundwasser Consulting GmbH and engineering consultants Beger for environmental analysis and research, can also be used as a commercially operated service system, for example to carry out initial analyses for approval procedures, to create measure plans, or to train task forces. The pilot prototype of the system can be quickly modified and enhanced to meet the special requirements of customers – for example chemical companies or fire departments.
For ERAMAS, the team used the technology of the Fraunhofer Resource Grid (FhRG), which is based on the Globus Toolkit. This software package includes various services that enable standardized access to distributed resources. With the help of the FhRG technology, the individual ERAMAS software and hardware components – such as the simulation module for the ground and air dispersion of pollutants and the computing capacities distributed across organizations – which are localized at project partners, can be linked in an SOA over the internet. This is achieved by encapsulating them on the basis of a component model. Special grid software monitors the hardware resources, optimizes the load, and provides additional grid functionalities as web services. Possible workflows are available in the form of workflow descriptions, and can be reliably executed on the grid as needed with the help of a workflow service.
Complex simulations in minutes
The system’s strength lies in its flexibility. It selects the most suitable model and system information for the problem at hand. In this way, it takes account of the underlying conditions of a particular environmental problem, for example in relation to the affected area, or the type of substance released into the atmosphere. If large volumes of a toxic substance evaporate during an accident, ERAMAS first determines the weather data for the accident site and installs a suitable evaporation model on the available high-performance computers. This estimates the most probable and the maximum evaporation rate to be expected depending on the substance parameters, clouds, wind speed, and temperature. At the same time, the system uses the data from a weather station and the height structures (orography) of the area to calculate a three-dimensional wind field. The evaporation rate and wind field are the input data for a dispersion model, which determines the pollutant concentrations over time for all points of the affected area. These results are in turn linked to models that forecast the infiltration in the soil and the transportation of pollutants in the groundwater and evaluate the effects on the environment and human health.
Thanks to grid technology, the complete workflow – from the input of accident data through to the evaluation of short-term and long-term dangers for people and the environment – can be reduced to just a few minutes. As ERAMAS is designed as a distributed system the models of which are managed by developers, new scientific or technical developments such as new simulation areas or additional transport paths can be integrated easily.
K-Wf Grid: Clear view through the service jungle
The grid of the future will contain a wide variety of services, some of which will build cooperatively on each other – orthogonal services that supplement each other – while some of which will compete against each other – services offering the same or similar functionalities from different providers. To be able to effectively use an environment of this complexity, a supporting infrastructure is required. This is where the K-Wf Grid (Knowledge-based Workflow System for Grid Applications) project comes in. The project involves research institutes and companies from various countries in the European Union, and is led by Fraunhofer FIRST. The aim is to develop an expert system that collects and provides knowledge over the grid, for example which workflow was used most often to successfully solve a problem.
K-Wf Grid helps users to select services and create complex workflows for grid applications in a service-oriented architecture. Software components that offer functionalities as services in the grid can be combined in different ways to form larger applications that can be executed effectively. To this end, the system monitors all interactions, for example the execution of applications on the grid, and evaluates these. As a result, a large base of data is available that can be used to develop workflows and guarantee the quality of the services. The execution time, reliability and quality of the results of a linked application can thus be estimated and optimized prior to execution.
The system is currently being enhanced with specific application knowledge, including the conditions under which particular programs can be used and linked, and the meaning and scope of input data. On the basis of this knowledge, the system is able to make suggestions to users for tried-and-tested workflows, for example.
The effectiveness of the K-Wf Grid is revealed in its use with three different applications: Together with the “Slovak Water Management Enterprise,” simulations on flooding predictions have been achieved in which – as in ERAMAS – different environmental simulation programs are made available as services and combined to form a workflow. Second, a traffic management system based on the K-Wf Grid was developed for the city of Genoa in Italy. This system forecasts the effect that blockages, construction sites, and traffic congestion have on the flow of traffic and the environmental stress of the city. Third, the company LogicDIS S.A. ported its client-server-based ERP system on the service-oriented K-Wf Grid in order to optimize the execution of processes in an ERP system.
MediGRID: Basic medical research
The MediGRID joint initiative uses the example of biomedical research to show how grid services can be applied in medicine and the life sciences. The project consortium, led by the umbrella organization TMF e.V., consists of representatives from biomedical research such as the universities and university clinics of Göttingen, Marburg, Konstanz, Dresden, and the Berliner Charité, on the one hand. On the other, the Fraunhofer Institute for Industrial Engineering (IAO) and FIRST, along with the Berlin-based Zuse Institute, contribute their proven skills in the areas of grid computing and distributed data management.
The MediGRID pilot application “RNAi Pipeline” is used to predict the functional properties of proteins of unknown 3D structure. Starting with the protein sequences, the system uses a comparison with known, classified structures to propose protein structure models (fold recognition). With this information, the application first analyzes the interaction with other biomolecules, known as pathways. As a large number of sequences of this kind need to be investigated independently, the complete processing pipeline, consisting of three main components, can run in parallel for the sequences.
The previous implementation of the application had already been based on an SOA – the main components are designed as static web services, and communicate via SOAP. However, this does not allow the potential for parallel processing described above to be exploited effectively, and the mere replication of services is not suitable for the dynamic job volume and generates considerable additional organizational and administrative workload. In the MediGRID project, the application is therefore integrated into a Globus Toolkit 4-based community grid for biomedical research in Germany, to allow additional computer resources to be integrated on an as-needed basis.
By migrating components of the RNAi pipeline into the grid services, it is possible to assign the services required in a way that saves the resources. The middleware and resource description developed in the Fraunhofer Resource Grid ensures that the system selects the resources that are best suited to executing the service and are not currently fully utilized, from the pool of available computing resources.
In addition to the benefits relating to interoperability and security already offered by a service-oriented basic architecture, grid integration enables resources to be virtualized to a large extent. As a result, it is possible to take full advantage of the parallelism inherent to the application problem on the available resources without generating additional administrative workload. Simplified, web-based access to the application via the shared grid portal of the MediGRID project is currently in development.